Lost in Distillation: A Case Study in Toxicity Modeling
Alyssa Chvasta, Alyssa Lees, Jeffrey Sorensen, Lucy Vasserman, Nitesh Goyal
Abstract
In an era of increasingly large pre-trained language models, knowledge distillation is a powerful tool for transferring information from a large model to a smaller one. In particular, distillation is of tremendous benefit when it comes to real-world constraints such as serving latency or serving at scale. However, a loss of robustness in language understanding may be hidden in the process and not immediately revealed when looking at high-level evaluation metrics. In this work, we investigate the hidden costs: what is “lost in distillation”, especially in regards to identity-based model bias using the case study of toxicity modeling. With reproducible models using open source training sets, we investigate models distilled from a BERT teacher baseline. Using both open source and proprietary big data models, we investigate these hidden performance costs.- Anthology ID:
- 2022.woah-1.9
- Volume:
- Proceedings of the Sixth Workshop on Online Abuse and Harms (WOAH)
- Month:
- July
- Year:
- 2022
- Address:
- Seattle, Washington (Hybrid)
- Venue:
- WOAH
- SIG:
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 92–101
- Language:
- URL:
- https://aclanthology.org/2022.woah-1.9
- DOI:
- 10.18653/v1/2022.woah-1.9
- Cite (ACL):
- Alyssa Chvasta, Alyssa Lees, Jeffrey Sorensen, Lucy Vasserman, and Nitesh Goyal. 2022. Lost in Distillation: A Case Study in Toxicity Modeling. In Proceedings of the Sixth Workshop on Online Abuse and Harms (WOAH), pages 92–101, Seattle, Washington (Hybrid). Association for Computational Linguistics.
- Cite (Informal):
- Lost in Distillation: A Case Study in Toxicity Modeling (Chvasta et al., WOAH 2022)
- PDF:
- https://preview.aclanthology.org/ingestion-script-update/2022.woah-1.9.pdf
- Data
- C4, WikiConv